A new study published in Nature Human Behaviour analyzed web-traffic data collected during the 2016 US presidential campaign. The study found that Trump supporters were most likely to consume news from untrustworthy sources but that, in general, untrustworthy news took up only around 6% of the total news consumed by Americans.
Researchers Andrew M. Guess and his team wanted to investigate how the consumption of news from unreliable sources might influence political behavior. “Fake news remains one of the most widely debated aspects of the 2016 US presidential election,” the researchers say. “Some journalists and researchers have even suggested that fake news may be responsible for Trump’s victory.”
To investigate these ideas, the study authors analyzed pre-election survey data collected between October 21-31, 2016 from a sample of 2,525 Americans. This survey data was then compared to the web-traffic histories of respondents, which was recorded from October 7 to November 14, 2016. The researchers used Grinberg et al.’s (2015) classification of fake news to identify websites that were either flagged by fact-checkers for publishing false content or flagged by human review for a flawed editorial process.
Researchers tracked respondents’ visits to hard-news sites — websites with a focus on “national news, politics, or world affairs.” They then separately calculated visits to websites classified as “untrustworthy”. Finally, the researchers calculated what proportion of the total news consumed by respondents came from these unreliable sites.
Results suggested that exposure to so-called “fake news” may be less prevalent than many have speculated. First, less than half of the sample (44%) had visited an untrustworthy news site during the time of the study. Next, news articles from untrustworthy sites made up only about 6% of all the hard-news articles read by Americans during this time.
Still, interesting trends emerged suggesting that certain groups were more likely to read fake news. Results showed that as much as 62% of traffic to untrustworthy websites came from the consumers who ranked in the top 20% for “most conservative information diet.” The authors explain, “people who indicated in the survey that they supported Trump were far more likely to visit untrustworthy websites—especially those who are conservative and are therefore probably pro-Trump—compared with those who indicated that they were Clinton supporters.”
It also appeared that subjects were more likely to consume news from unreliable sources when its content corresponded to their political beliefs. When it came to articles from untrustworthy conservative websites, 57% of Donald Trump supporters had read at least one of these articles, while only 28% of Hillary Clinton supporters had. When it came to articles from unreliable liberal websites, 23% of Clinton supporters had read at least one of these, while only 11% of Trump supporters had.
Next, the researchers’ analysis suggested that social media, and in particular, Facebook, might serve as a gateway to untrustworthy news. Evidence suggested that the traffic to untrustworthy sites often stemmed from the social media platform. “Facebook was among the three previous websites visited by respondents in the previous 30 s for 15.1% of the articles from untrustworthy news websites that we observed in our web data,” the authors say.
When it came to predicting political behavior, equivalency tests ruled out large effects for the consumption of untrustworthy news on voter choice or turnout. Still, the authors express that their results were too imprecise to say with certainty whether or not untrustworthy news may have aided Trump’s win.
The authors express that although exposure to fake news appears to be limited to a specific subset of Americans, “these small groups can help to propel dubious claims to widespread visibility online, potentially intensifying polarization and negative affect.” The researchers express that it would be interesting for future research to further explore these effects by including news accessed through social media platforms directly, such as “hyper-partisan” Facebook groups and Twitter feeds.
The study, “Exposure to untrustworthy websites in the 2016 US election”, was authored by Andrew M. Guess, Brendan Nyhan, and Jason Reifler.